The hybrid silicon photonics platform: 10 years of innovation

[Curt_Hopkins]

Hewlett Packard Labs’ new hybrid silicon platform is the result of a decade-plus of exacting labor and moments of punctuated intuition. This braid of innovation and hard work was impressive enough to the IEEE that they requested the researchers submit a paper on the process and parameters involved.

That paper, “An Energy-Efficient and Bandwidth-Scalable DWDM Heterogeneous Silicon Photonics Integration Platform,” has been published in the organization’s Journal of Selected Topics in Quantum Electronics, Volume 28, Issue 6 November/December 2022.

“The paper is a summary of building an energy-efficient optical link using all the technology that we've been working on for the last 10 to 15 years,” said Geza Kurczveil, senior research scientist at the Santa Barbara-based Large-Scale Integrated Photonics Lab (LSIP).

It explains the work that went into transitioning from a VCSEL-based architecture that saw light coupled directly into a fiber to a new link that allows developers to interpose whatever material they wish between the source of light and its destination.

The earlier architecture structured featured no silicon.

“Those devices worked very well,” said Kurczveil. “VCSELs are dirt cheap, and they work absolutely fantastically. But the issue is scaling. Every time you want to improve or increase the bandwidth, your packaging cost goes up. But with the technology that we have on silicon photonics, it does not matter how many channels you have, the packaging cost is fixed because you just attach a fiber to your output waveguide.”

Another selling point for the hybrid silicon photonics platform is how well it handles an increasingly important reality of modern life, one that has started to change the very quality of our computing – big data.

“Whenever you have a massive amount of data, you want a high bandwidth link, and that is what we are aiming for,” said Kurczveil. “You have lasers that emit multiple wavelengths at once. And we are in the process of scaling up the number of channels.”

So the hybrid platform has enabled devices that emit something on the order of 40 simultaneous channels.

“If we can scale that up to 128 channels that's going to give you a load of bandwidth,” he said. With that amount of data moving around, you want to do it using the least amount of energy possible. The avalanche photo detectors detect lower power signals so you can drive your laser at a lower drive current. The same goes for MOS-based components. Without MOS, people would use thermal tuners to align their ring resonance to a comb laser, or to tune the interleavers to compensate for fabrication imperfections and thermal tuners can consume a lot of power. With Labs’ MOS-based tuning six orders of magnitude less power is used.

Something people outside research may not realize, and that rarely comes across in publications, is that the culture of a lab has a real tangible effect on how its researchers work and that was certainly the case with the LSIP. The fact that its researchers call themselves “Santa Barbarians” is not inconsequential to the way they work. The “barbarism” in question describes is a ferocious favoring of facts.

“We are researchers first, so that means that the truth is more important than any opinion,” said Kurczveil. “It doesn't matter who says the truth. The truth will always rule, absolutely. If leadership has an opinion and an intern says, ‘You're wrong and here's the data,’ the senior researcher loses. That's just how it goes. That means that we don't really have egos. We are researchers, we focus on the thing that actually matters.”

This work is getting people’s attention because the researchers are combining state of the art research on three completely different topics – comb lasers, MOS modulation/tuning, and avalanche photodetectors – and applying it to one application, high performance computing.

“Most researchers don’t work on two things at once, let alone three,” says Kurczveil. “Even then the work consists mostly of incremental improvements, rather than radically new ideas. We combine multiple paths of research with a thirst for radical change. I think people are rooting for this technology to be part of post-exascale systems.”

Curt Hopkins
Hewlett Packard Enterprise

twitter.com/hpe_labs
linkedin.com/showcase/hewlett-packard-labs/
labs.hpe.com